Free a build node's disk after every build and refuse work under a floor of free space (issue 1834)

A full / made every check on shanks end 'store never became ready' with the reason erased by --rm.
Builds now remove their local image tag, narrowed source and context clone; a build or check
refuses under 20 GiB free on / and the workspace, saying so; the throwaway store is no longer
started with --rm and a store that does not come up reports its exit code and last 20 log lines.
This commit is contained in:
2026-10-22 13:14:00 +02:00
parent 4463991e06
commit 995f8237ce
6 changed files with 322 additions and 7 deletions
+39 -3
View File
@@ -392,6 +392,12 @@ func Check(ctx context.Context, run Runner, spec CheckSpec, workspace, registry
ctx, stop := context.WithTimeout(ctx, CheckTimeout)
defer stop()
// **A node with no room takes no check** (novox/hq issue 1834): said as the check's verdict, never a throwaway store
// that dies at once on a full disk.
if err := RoomToWork("/", workspace); err != nil {
say("check", "refused: %v", err)
return CheckVerdict{}, err
}
root := filepath.Join(workspace, "check")
if err := os.RemoveAll(root); err != nil {
return CheckVerdict{}, err
@@ -1631,7 +1637,9 @@ func BusImage(version string) string {
// throwaway starts a container publishing one port on loopback, and answers where it is reached.
func throwaway(ctx context.Context, run, plain Runner, name, image string, port int, opts, args []string) (string, error) {
invocation := []string{"run", "-d", "--rm", "--name", name, "-p", fmt.Sprintf("127.0.0.1::%d", port)}
// Never with --rm (novox/hq issue 1834): a store that dies at once took its log with it. The check removes what it
// raised, with its volumes, after it has read why one stopped (RemoveContainersOf).
invocation := []string{"run", "-d", "--name", name, "-p", fmt.Sprintf("127.0.0.1::%d", port)}
invocation = append(invocation, opts...)
invocation = append(invocation, image)
invocation = append(invocation, args...)
@@ -1652,17 +1660,45 @@ func throwaway(ctx context.Context, run, plain Runner, name, image string, port
// waitFor runs a readiness command in a container until it answers, for a minute.
func waitFor(ctx context.Context, run Runner, name string, ready []string) error {
for i := 0; i < 60; i++ {
return waitReady(ctx, run, name, ready, 60)
}
// waitReady is waitFor for a number of tries, a second apart. A container that never answers is not left a mystery
// (novox/hq issue 1834): its state and exit code and the last 20 lines of its log are in the error, read before the
// check removes it.
func waitReady(ctx context.Context, run Runner, name string, ready []string, attempts int) error {
for i := 0; i < attempts; i++ {
if _, err := run(ctx, "", "docker", append([]string{"exec", name}, ready...)...); err == nil {
return nil
}
if i == attempts-1 {
break
}
select {
case <-ctx.Done():
return ctx.Err()
case <-time.After(time.Second):
}
}
return fmt.Errorf("%s never became ready", name)
return fmt.Errorf("%s never became ready%s", name, whyItStopped(ctx, run, name))
}
// whyItStopped is ": <state>, exit code N; its last log lines: …" for a container, or "" when docker says nothing.
func whyItStopped(ctx context.Context, run Runner, name string) string {
reading, done := context.WithTimeout(context.WithoutCancel(ctx), 30*time.Second)
defer done()
var said []string
if out, err := run(reading, "", "docker", "inspect", "--format",
"{{.State.Status}}, exit code {{.State.ExitCode}}", name); err == nil && strings.TrimSpace(out) != "" {
said = append(said, strings.TrimSpace(out))
}
if out, err := run(reading, "", "docker", "logs", "--tail", "20", name); err == nil && strings.TrimSpace(out) != "" {
said = append(said, "its last log lines: "+strings.TrimSpace(out))
}
if len(said) == 0 {
return ""
}
return ": " + strings.Join(said, "; ")
}
// dialable waits for an address to take a connection, for a minute.